Last week, the U.S. Food and Drug Administration (FDA) released its discussion paper, “Proposed Regulatory Framework for Modifications to Artificial Intelligence/Machine Learning (AI/ML)-Based Software as a Medical Device (SaMD).” The purpose of the document is to present preliminary concepts and generate public feedback to inform future policy approaches to total product lifecycle oversight of AI/ML SaMD.
The focus is on SaMD modified or adaptable through AI/ML-based methodologies. To date, developers of AI/ML-informed SaMD have “locked” updates to their algorithms due to uncertainty and perceived complexity of FDA oversight of post-market modifications to AI/ML SaMD. The discussion paper addressed various categories of modifications that could alter algorithm architecture or involve re-training with new data sets, such as clinical and analytical performance changes, input differences or intended use updates.
The paper covers a possible framework that gives manufacturers the option to submit certain types of plans for foreseeable modifications during the FDA’s initial premarket review of the AI/ML SaMD. The agency would review those plans to essentially pre-vet future changes. The paper also addresses post-market scenarios in which manufacturers may need to deviate from their modification plans.
The concepts in the discussion paper borrow from multiple existing FDA regulatory pathways and agency-led initiatives including the International Medical Device Regulators Forum risk categorization principles for SaMD, the Digital Health Software Pre-Certification Program, and the 510(k), De Novo and Premarket Approval pathways. The FDA noted, however, that legislation may be required to expand the agency’s authority to allow future implementation of the framework.
Members interested in providing feedback on this discussion paper or other FDA Digital Health initiatives should contact Michael Peters, ACR Director of Legislative and Regulatory Affairs, at firstname.lastname@example.org.